2.2.6 -> 2.3.1 failed



  • This past Friday, I tried to upgrade my trusty old 2.2.6 install to 2.3.1.  The VM 2.2.6 was living on was old and had been upgraded several times, with small hacks like the Sarg symbolic link fix etc applied and I wanted to start clean so I created a new VM with vmx NICs to replace the em NICs in the old VM.  I thought it would be a simple matter of exporting my backup config and replacing em with vmx for WAN, LAN and OPT1, and then re-import into the new VM.  When I did that I had no Internet access on bootup and the rest of the upgrade barfed.

    Next, I decided to try it with the existing 2.2.6 box.  I ran the GUI auto-upgrader and everything seemed to work right.  However, after it rebooted squid did not run no matter what I tried in the short time I had.  This is important for this node so I ended up reverting to my snapshot of 2.2.6.

    I only have a small maintenance window Friday 5:00-5:30 so I have to get my act together as much as possible beforehand.  I could debug the squid issue after upgrading from 2.2.6 but I would prefer to start on a clean VM and I can't get Internet access to download the packages during upgrade.

    Any suggestions as to why the new VM with vmx NICs wouldn't work when importing a modified config would be appreciated.


  • LAYER 8 Global Moderator

    Well squid was changed.. Removed and new package created

    https://doc.pfsense.org/index.php/2.3_Removed_Packages
        squid / squid3 - the squid packages have been consolidated into a single squid package, using version 3.5. The "squid3" package has been removed, and the "squid" package is now version 3.

    squidGuard / squidGuard-devel - the squidGuard packages have been consolidated into a single option. The squidGuard-devel version no longer exists.

    So could see possible issues related to that causing you your pain.

    As to config from 2.6 using e1000 being applied to clean 2.3.1 with vmx - yeah going to have remap all your interfaces for sure with that sort deployment.  Why not try that with using e1000, on your clean install.  Then adding your vmx interfaces and remapping your interfaces.

    Your best bet would prob be to remove all your packages.  Do your upgrade, get your nics working with vmx if that is what you want.  Personally I still use the 1000 because the lvad package that does lldp and cdp when using vmx my switch log floods with dup mismatch errors.  But no such problems with e1000, have yet to see any real significant performance improvements using vmx over e1000.  Maybe if your host was running lots and lots of vms, etc.  I do use the vmx on my other vms, just on pfsense just still use the e1000 vnics.

    Then after up and running you can add your packages you want that have been brought into 2.3

    What I can tell you is upgraded my 2.2.6 to 2.3rc, then 2.3 full, the tiny update 2.3_1 and now 2.3.1 without any issues at all.  Packages I use were not any of the problem packages, or changed packages, etc. Running on esxi6



  • The base layer always works, it's the extras that seem to have issues.  I can work through the squid stuff as I'm fairly proficient in that area.  I was just hoping that I could import my config without problems and move on from there.  I'll try again with the config import this Friday with em NICs on a clean VM.


Log in to reply